How do you teach DoE? - Practical Example Discussion

[#1]

Hello community!

I recently saw a great LinkedIn post from Morten Nielsen on a 3D printed 'black box' tool to practically teach to complete DoE novices and it got me thinking about what examples work best to teach DoE. I'm sure a lot of us have been through the paper helicopter example in our time, which is a great (and more importantly, fun!) example and pairs nicely with the Prediction Profiler tool - but it uses mainly categorical factors and can take a long time for novices to perform.

What examples have you used that are quick, able to show complex problems and most of all get people interested in DoE?

Cheers!

Ben

[#2 · Accepted solution]

Hi @Ben_Ingham ,

Thank you for the sharing.

I used material from JMP community for our internal training (this example is quite simple for beginner, because I can re-use the value to fit our scenario: i.e change the factors/response name)

https://community.jmp.com/kvoqx44227/attachments/kvoqx44227/learn-jmp-tutorials-tkb/388/5/EZ%20DOE%2... 

p/s: Most of our engineers only use DOE platform to use prediction profiler tool.

[#3 · Accepted solution]

My 'go to' experiential DOE learning tools were the paper helicopters and some version of the 'Statapult'. Back in my day of training Eastman Kodak Company Six Sigma Black Belts we used a custom designed and fabricated statapult that had a combination of discrete numeric (eg. number of rubber bands), continuous ( eg. arm length), and categorical factors (eg. ball type, like foam, solid plastic, plastic with holes). The team had to design their own measurement system. People did things like build a sand filled landing pit (think long jump pit) or two people estimating landing point along a tape measure, average the observations. The statapult was a great tool because it so much emulated real life in the lab or production. People got confused, it broke, nuisance variables galore, you name it it happened. One participant moaned about all the operational issues...another on their team told them something like, '...this is what happens in real life...learning how deal with these issues is half of what we're learning here...not just statistics'.

Another, more specialized choice was in our Measurement Systems Analysis module. At the end of the module we had teams set up a typical MSA study...which is still a designed experiment, just with special characteristics like nesting, crossing, replication, operators, devices, and on and on plus specialized analysis methods...which we focused on Wheeler's EMP approach. The teams were given a response...which was the 'length in inches of a plastic straw'. They were given multiple devices to assess length. Things like a plastic ruler, calipers, a micrometer, etc. Then told to define a measurement system and analyze the results using Wheeler's method. But we threw a few monkey wrenches at them. There were 4 or 5 straws that were the 'parts'. One straw had a beveled edge...so there was always lots of conversation about how to actually measure length...when one end of the straw has two distinct end points. One straw was too long to fit in the jaws of the caliper. One 'device' was delineated in metric...so they had to do a conversion to inches. This module was always lots of fun to watch the participants work through all the issues.

[#4 · Accepted solution]

In my classes, I have the participants run several hands-on experiments (they start with basic (tablet dissolution rate) to complex (optimizing the flight of a balsa airplane) requiring multiple iterations).  These are both educational and entertaining.  I also require students to apply the methodology to their work environment.  They must design multiple experiments on projects they are working on (not necessarily run them).  In addition, I have them run an experiment on something they enjoy personally.  I give them Bill Hunter's "101 DOE Ideas" paper and "DOE It Yourself Fun science projects compiled by Mark J. Anderson, Principal, Stat-Ease, Inc.